def convert(values, unit1, unit2):
"""Converts values from unit1 to unit2
Parameters
----------
values: ndarray
Values of the field to convert
unit1: str
start unit
unit2: str
final unit
Returns
-------
ndarray of the converted field
"""
unit1_save = unit1
unit2_save = unit2
# Format the strings
unit1 = unit1.replace("*", "").replace(" ", "").replace("^", "")
unit2 = unit2.replace("*", "").replace(" ", "").replace("^", "")
# Unit1 parsing
if "/" in unit1:
dim1_denom, prefix1_denom = get_dim_prefix(unit1.split("/")[1])
unit1 = unit1.split("/")[0]
else:
dim1_denom = [0, 0, 0, 0, 0, 0]
prefix1_denom = 1.0
dim1_num, prefix1_num = get_dim_prefix(unit1)
# Unit2 parsing
if "/" in unit2:
dim2_denom, prefix2_denom = get_dim_prefix(unit2.split("/")[1])
unit2 = unit2.split("/")[0]
else:
dim2_denom = [0, 0, 0, 0, 0, 0]
prefix2_denom = 1.0
dim2_num, prefix2_num = get_dim_prefix(unit2)
# Check compatibility
dim1 = [i - j for i, j in zip(dim1_num, dim1_denom)]
dim2 = [i - j for i, j in zip(dim2_num, dim2_denom)]
if dim1 != dim2:
raise UnitError("ERROR: Units " + unit1_save + " and " + unit2_save +
" do not match")
else:
return values * (prefix1_num / prefix1_denom) / (prefix2_num /
prefix2_denom)
def dB_to_dBA(values, freqs):
"""Converts values from dB into dBA (requires frequency vector)
Parameters
----------
values: array
Values of the field to convert (must be 1D)
freqs: array
Frequency vector
Returns
-------
ndarray of the converted field
"""
freq2 = square(freqs)
RA = (12200.0**2 * freq2**2 / ((freq2 + 20.6**2) * sqrt(
(freq2 + 107.7**2) * (freq2 + 737.0**2)) * (freq2 + 12200.0**2)))
Aweight = 2.0 + 20.0 * log10(RA)
try:
values += Aweight
return values
except:
raise UnitError("ERROR: dBA conversion only available for 1D fft")
def get_magnitude_along(
self, *args, unit="SI", is_norm=False, axis_data=[], is_squeeze=True
):
"""Returns the ndarray of the magnitude of the FT, using conversions and symmetries if needed.
Parameters
----------
self: Data
a Data object
*args: list of strings
List of axes requested by the user, their units and values (optional)
unit: str
Unit requested by the user ("SI" by default)
is_norm: bool
Boolean indicating if the field must be normalized (False by default)
axis_data: list
list of ndarray corresponding to user-input data
Returns
-------
list of 1Darray of axes values, ndarray of magnitude values
"""
if len(args) == 1 and type(args[0]) == tuple:
args = args[0] # if called from another script with *args
return_dict = self.get_along(
args, axis_data=axis_data, is_squeeze=is_squeeze, is_magnitude=True
)
values = return_dict[self.symbol]
# 1/nth octave band
for axis in return_dict["axes_list"]:
if axis.name == "freqs" or axis.corr_name == "freqs":
index = axis.index
if axis.noct is not None:
(values, foct) = apply_along_axis(
to_noct, index, values, return_dict["freqs"], noct=axis.noct
)
return_dict[axis.name] = foct
# Convert into right unit (apart because of dBA conversion)
if unit == self.unit or unit == "SI":
if is_norm:
try:
values = values / self.normalizations.get("ref")
except:
raise NormError(
"ERROR: Reference value not specified for normalization"
)
elif unit == "dB":
ref_value = 1.0
if "ref" in self.normalizations.keys():
ref_value *= self.normalizations.get("ref")
values = to_dB(values, self.unit, ref_value)
elif unit == "dBA":
ref_value = 1.0
if "ref" in self.normalizations.keys():
ref_value *= self.normalizations.get("ref")
if "freqs" in return_dict.keys():
for axis in return_dict["axes_list"]:
if axis.name == "freqs" or axis.corr_name == "freqs":
index = axis.index
values = apply_along_axis(
to_dBA, index, values, return_dict["freqs"], self.unit, ref_value
)
else:
raise UnitError(
"ERROR: dBA conversion only available for fft with frequencies"
)
elif unit in self.normalizations:
values = values / self.normalizations.get(unit)
else:
values = convert(values, self.unit, unit)
return_dict[self.symbol] = values
return return_dict
def dB_to_dBA(values, freqs, noct=None):
"""Converts values from dB into dBA (requires frequency vector)
Parameters
----------
values: array
Values of the field to convert (must be 1D)
freqs: array
Frequency vector
Returns
-------
ndarray of the converted field
"""
if noct is not None:
freqs_ref = [
12.5,
16,
20,
25,
31.5,
40,
50,
63,
80,
100,
125,
160,
200,
250,
315,
400,
500,
630,
800,
1000,
1250,
1600,
2000,
2500,
3150,
4000,
5000,
6300,
8000,
10000,
12500,
16000,
20000,
]
Aweight = [
-63.4,
-56.7,
-50.5,
-44.7,
-39.4,
-34.6,
-30.2,
-26.2,
-22.5,
-19.1,
-16.1,
-13.4,
-10.9,
-8.6,
-6.6,
-4.8,
-3.2,
-1.9,
-0.8,
0,
0.6,
1,
1.2,
1.3,
1.2,
1,
0.5,
-0.1,
-1.1,
-2.5,
-4.3,
-6.6,
-9.3,
]
mask = argwhere(freqs in freqs_ref)
Aweight = take(Aweight, mask)
else:
freq2 = square(freqs)
freq2[freq2 == 0] = nan
RA = (
12200.0 ** 2
* freq2 ** 2
/ (
(freq2 + 20.6 ** 2)
* sqrt((freq2 + 107.7 ** 2) * (freq2 + 737.0 ** 2))
* (freq2 + 12200.0 ** 2)
)
)
Aweight = 2.0 + 20.0 * log10(RA)
Aweight[isnan(Aweight)] = -100 # replacing NaN by -100 dB
Aweight[
values <= 0
] = 0 # avoiding to increase dB in dBA at frequencies where noise is already null
try:
values += Aweight
return values
except:
raise UnitError("ERROR: dBA conversion only available for 1D fft")
def convert(self, values, unit, is_norm, is_squeeze, is_magnitude, axes_list):
"""Returns the values of the field transformed or converted.
Parameters
----------
self: Data
a Data object
values: ndarray
array of the field
unit: str
Unit requested by the user ("SI" by default)
is_norm: bool
Boolean indicating if the field must be normalized (False by default)
Returns
-------
values: ndarray
values of the field
"""
# Take magnitude before summing
if is_magnitude:
values = np_abs(values)
# Apply sums, means, etc
for axis_requested in axes_list:
# sum over sum axes
if axis_requested.extension == "sum":
values = np_sum(values, axis=axis_requested.index, keepdims=True)
# root sum square over rss axes
elif axis_requested.extension == "rss":
values = sqrt(
np_sum(values**2, axis=axis_requested.index, keepdims=True))
# mean value over mean axes
elif axis_requested.extension == "mean":
values = np_mean(values, axis=axis_requested.index, keepdims=True)
# RMS over rms axes
elif axis_requested.extension == "rms":
values = sqrt(
np_mean(values**2, axis=axis_requested.index, keepdims=True))
# integration over integration axes
elif axis_requested.extension == "integrate":
values = trapz(
values, x=axis_requested.values,
axis=axis_requested.index) / (np_max(axis_requested.values) -
np_min(axis_requested.values))
if is_squeeze:
values = squeeze(values)
if unit == self.unit or unit == "SI":
if is_norm:
try:
values = values / self.normalizations.get("ref")
except:
raise NormError(
"ERROR: Reference value not specified for normalization")
elif unit == "dB":
ref_value = 1.0
if "ref" in self.normalizations.keys():
ref_value *= self.normalizations.get("ref")
values = to_dB(np_abs(values), self.unit, ref_value)
elif unit == "dBA":
ref_value = 1.0
if "ref" in self.normalizations.keys():
ref_value *= self.normalizations.get("ref")
for axis in axes_list:
is_match = False
if axis.name == "freqs" or axis.corr_name == "freqs":
index = axis.index
values = apply_along_axis(to_dBA, index, values, axis.values,
self.unit, ref_value)
is_match = True
elif axis.name == "frequency":
index = axis.index
values = apply_along_axis(to_dBA, index, values, axis.values,
self.unit, ref_value)
is_match = True
if not is_match:
raise UnitError(
"ERROR: dBA conversion only available for fft with frequencies"
)
elif unit in self.normalizations:
values = values / self.normalizations.get(unit)
else:
values = convert_unit(values, self.unit, unit)
return values
def get_harmonics(self,
N_harm,
*args,
unit="SI",
is_norm=False,
is_flat=False):
"""Returns the complex Fourier Transform of the field, using conversions and symmetries if needed.
Parameters
----------
self: Data
a Data object
N_harm: int
Number of largest harmonics to be extracted
args: list
Axes names, ranges and units
unit: str
Unit demanded by the user ("SI" by default)
is_norm: bool
Boolean indicating if the field must be normalized (False by default)
is_flat: bool
Boolean if the output data remains flattened (for 2D cases)
Returns
-------
list of 1Darray of axes values, ndarray of magnitude of FT
"""
# Read the axes input in args
if len(args) == 1 and type(args[0]) == tuple:
args = args[0] # if called from another script with *args
axes_list = read_input_strings(args, [])
# Extract the requested axes (symmetries + unit)
for axis_requested in axes_list:
if axis_requested[3] == "values":
# Get original values of the axis
axis_requested.append(
self.get_FT_axis(axis_requested[0] + axis_requested[1]))
# Interpolate axis with input data
if str(axis_requested[4]) == "whole":
axis_requested[4] = axis_requested[5]
axis_requested[5] = "whole"
else:
axis_requested[4] = get_common_base(axis_requested[5],
axis_requested[4])
# Change fft name for the slices of the field
if axis_requested[0] == "freqs":
axis_requested[0] = "time"
elif axis_requested[0] == "wavenumber":
axis_requested[0] = "angle"
# Check if the requested axis is defined in the Data object
for axis_requested in axes_list:
axis_name = axis_requested[0]
is_match = False
for index, axis in enumerate(self.axes):
if axis.name == axis_name:
is_match = True
if not is_match:
axes_list.remove(axis_requested)
# Rebuild symmetries of field if axis is extracted
values = self.values
for index, axis in enumerate(self.axes):
for axis_requested in axes_list:
if axis.name in self.symmetries.keys(
) and axis.name == axis_requested[0]:
values = rebuild_symmetries(values, index,
self.symmetries.get(axis.name))
# Extract the slices of the field (single values)
for index, axis in enumerate(self.axes):
is_match = False
for axis_requested in axes_list:
if axis.name == axis_requested[0]:
is_match = True
if axis_requested[3] == "indices" and axis_requested[
2] == "single":
values = take(values, axis_requested[4], axis=index)
if not is_match: # Axis was not specified -> take slice at the first value
values = take(values, [0], axis=index)
# Interpolate over axis values (single values)
for index, axis in enumerate(self.axes):
for axis_requested in axes_list:
if (axis.name == axis_requested[0]
and axis_requested[3] == "values"
and axis_requested[2] == "single"):
values = apply_along_axis(
get_interpolation,
index,
values,
axis_requested[5],
axis_requested[4],
)
# Perform Fourier Transform
values = np_abs(comp_fft(values))
# Extract slices again (intervals)
for index, axis in enumerate(self.axes):
for axis_requested in axes_list:
if axis.name == axis_requested[0]:
if axis_requested[2] == "indices" and axis_requested[
2] == "interval":
values = take(values, axis_requested[4], axis=index)
# Interpolate over axis values again (intervals)
for index, axis in enumerate(self.axes):
for axis_requested in axes_list:
if axis.name == axis_requested[0]:
if axis_requested[3] == "values" and axis_requested[
2] == "interval":
values = apply_along_axis(
get_interpolation,
index,
values,
axis_requested[5],
axis_requested[4],
)
# Eliminate dimensions=1
values = squeeze(values)
# Test if data is 1D or 2D
if len(values.shape) > 2:
raise AxisError("ERROR: only 1D or 2D implemented")
else:
# Convert into right unit
if unit == self.unit or unit == "SI":
if is_norm:
try:
values = values / self.normalizations.get("ref")
except:
raise NormError(
"ERROR: Reference value not specified for normalization"
)
elif unit == "dB":
ref_value = 1.0
if "ref" in self.normalizations.keys():
ref_value *= self.normalizations.get("ref")
values = to_dB(values, self.unit, ref_value)
elif unit == "dBA":
ref_value = 1.0
is_match = False
if "ref" in self.normalizations.keys():
ref_value *= self.normalizations.get("ref")
for axis_requested in axes_list:
if axis_requested[0] == "time":
is_match = True
values = to_dBA(values, axis_requested[4], self.unit,
ref_value)
if not is_match:
raise UnitError(
"ERROR: dBA conversion only available for fft with frequencies"
)
elif unit in self.normalizations:
values = values / self.normalizations.get(unit)
else:
values = convert(values, self.unit, unit)
# 1D case
if len(values.shape) == 1:
for axis_requested in axes_list:
if axis_requested[2] == "interval":
axis_values = axis_requested[4]
indices = argsort(negative(values))
indices = indices[:N_harm]
axis_values = axis_values[indices]
values = values[indices]
return [axis_values, values]
# 2D case
else:
for axis_requested in axes_list:
if axis_requested[0] == "angle":
r = axis_requested[4]
elif axis_requested[0] == "time":
f = axis_requested[4]
# Flatten the data
values_flat = values.flatten()
R, F = meshgrid(r, f)
f = F.flatten()
r = R.flatten()
# Get the N_harm largest peaks
indices = argsort(negative(values_flat))
indices = indices[:N_harm]
values = values_flat[indices]
f = f[indices]
r = r[indices]
if len(values.shape) == 2 and not is_flat:
f.reshape((N_harm, N_harm))
r.reshape((N_harm, N_harm))
values.reshape((N_harm, N_harm))
return [f, r, values]